From Points to Prints

Generation of building roofprints and footprints from ALS point clouds

Alexandre Bry

Context

Reconstruction of 3D buildings

  • Many applications of 3D buildings: visualization, simulations, solar potential, etc.
  • Different levels of detail required

Levels of detail

Visual example of the refined LoDs for a residential building (Biljecki et al., 2016).

Visual example of the refined LoDs for a residential building (Biljecki et al., 2016).

3DBAG example

Screenshot from the online viewer of the 3DBAG.

Screenshot from the online viewer of the 3DBAG.

French context

What about façade points?

  • ALS point clouds give high density of points on roofs and low density on façades

What characteristics of ALS point clouds influence the ability to generate accurate building roofprints and footprints?

Definitions

Building elements

Roof overhangs (eaves)

The portion of a roof that extends beyond the exterior walls of a building.

Roofprint

The vertical projection of all the roof components of a building.

Footprint

The vertical projection of the main façades of a building.

Illustrations

 

3D model from Sketchfab.

 

Footprint and roofprint (before vertical projection).

 

 

3D model from Sketchfab.

 

Footprint and roofprint (before vertical projection).

 

Figure 1: Examples of footprints (in blue) and roofprints (in red) for complex buildings.

Research context

Roof overhangs

Point cloud semantic segmentation

Datasets

Table 1: List of point cloud datasets for semantic segmentation of point clouds.
Name Number of points Point density (pts/m²) Classes for roof/façade Source Paper
H3D 74M 800 Yes ALS Kölle et al. (2021)
V3D 0.78M 8 Yes ALS Cramer (2010) Niemeyer et al. (2014)
DublinCity 260M 348 Yes ALS Zolanvari et al. (2019)
Campus3D 937.1M Yes Aerial images X. Li et al. (2020)
CENAGIS-ALS 550M 275 Yes ALS Zachar et al. (2023)
DALES 505M 50 No ALS Varney et al. (2020)
SensatUrban 2847M No Aerial images Hu et al. (2020)
LASDU 3.12M 4 No ALS Ye et al. (2020)
TALD 121M 12 No ALS Vijaywargiya & Ramiya (2025)
FRACTAL 9261M 37 No ALS Gaydon et al. (2024)

Methodology

Overview

  1. Train a machine learning model to isolate roof and façade points
  2. Use the segmented point cloud to compute roofprints
  3. Use the segmented point cloud and the roofprints to compute footprints
  4. Evaluate the method with generated data and hopefully real data

Semantic segmentation

  1. Generate a ground-truth dataset using 3DBAG and AHN4
    1. Download all data per tiles,
    2. Extend the roofs of buildings with fake overhangs,
    3. Compute the distances of each point to the set of roofs and to the set of façades,
    4. Classify the points based on these distances between roof, façade and other.
  2. Train a ML model on the data:
    1. Try different models (PointTransformerv3, KPConv, etc)
    2. Select and run the best one

LoD 2.2 buildings from the 3DBAG.

The classification of the points from AHN 4.
Figure 2: Example of classifying the AHN 4 with the 3DBAG.

Roofprints

  • The method still needs to be determined
  • It will depend on the results of semantic segmentation

Footprints

  • Sweep vertical planes parallel to the roofprint edges
  • Identify the best one or none at all
  • Potential to use symmetry of roofs

Assessment

  • Create a fake ground-truth dataset:
    • Create a fake LoD 2.3 dataset from 3DBAG LoD 2.2 by randomly extending the roof planes
    • Simulate an ALS point cloud with tools such as HELIOS++
  • Evaluate the method on this dataset
  • Some annotations of real data for better assessment?

Current and future work

Until now

  • Literature review
  • Explored French and Dutch datasets and their metadata
  • Started the work on classifying AHN 4 using 3DBAG

From now on

Project planning

Project planning

Thank you.

References

Biasutti, P., Bugeau, A., Aujol, J.-F., & Brédif, M. (2019). RIU-net: Embarrassingly simple semantic segmentation of 3D LiDAR point cloud. https://doi.org/10.48550/ARXIV.1905.08748
Biljecki, F., Ledoux, H., & Stoter, J. (2016). An improved LOD specification for 3D building models. Computers, Environment and Urban Systems, 59, 25–37. https://doi.org/10.1016/j.compenvurbsys.2016.04.005
Cramer, M. (2010). The DGPF-test on digital airborne camera evaluation overview and test design. Photogrammetrie - Fernerkundung - Geoinformation, 2010(2), 73–82. https://doi.org/10.1127/1432-8364/2010/0041
Dahlke, D., Linkiewicz, M., & Meissner, H. (2015). True 3D building reconstruction: Façade, roof and overhang modelling from oblique and vertical aerial imagery. International Journal of Image and Data Fusion, 6(4), 314–329. https://doi.org/10.1080/19479832.2015.1071287
Frommholz, D., Linkiewicz, M., Meissner, H., & Dahlke, D. (2017). Reconstructing buildings with discontinuities and roof overhangs from oblique aerial imagery. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLII-1/W1, 465–471. https://doi.org/10.5194/isprs-archives-xlii-1-w1-465-2017
Gaydon, C., Daab, M., & Roche, F. (2024). FRACTAL: An ultra-large-scale aerial lidar dataset for 3D semantic segmentation of diverse landscapes. https://doi.org/10.48550/ARXIV.2405.04634
Goebbels, S., & Pohle-Fröhlich, R. (2023). Automatic reconstruction of roof overhangs for 3D city models. Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, 145–152. https://doi.org/10.5220/0011604200003417
Hu, Q., Yang, B., Khalid, S., Xiao, W., Trigoni, N., & Markham, A. (2020). Towards semantic segmentation of urban-scale 3D point clouds: A dataset, benchmarks and challenges. https://doi.org/10.48550/ARXIV.2009.03137
Kölle, M., Laupheimer, D., Schmohl, S., Haala, N., Rottensteiner, F., Wegner, J. D., & Ledoux, H. (2021). The hessigheim 3D (H3D) benchmark on semantic segmentation of high-resolution 3D point clouds and textured meshes from UAV LiDAR and multi-view-stereo. ISPRS Open Journal of Photogrammetry and Remote Sensing, 1, 100001. https://doi.org/10.1016/j.ophoto.2021.100001
Li, M., Lin, S., Wang, Z., Shen, Y., Zhang, B., & Ma, L. (2024). Class-imbalanced semi-supervised learning for large-scale point cloud semantic segmentation via decoupling optimization. Pattern Recognition, 156, 110701. https://doi.org/10.1016/j.patcog.2024.110701
Li, W., Wang, F.-D., & Xia, G.-S. (2020). A geometry-attentional network for ALS point cloud classification. ISPRS Journal of Photogrammetry and Remote Sensing, 164, 26–40. https://doi.org/10.1016/j.isprsjprs.2020.03.016
Li, X., Li, C., Tong, Z., Lim, A., Yuan, J., Wu, Y., Tang, J., & Huang, R. (2020). Campus3D: A photogrammetry point cloud benchmark for hierarchical understanding of outdoor scene. Proceedings of the 28th ACM International Conference on Multimedia, 238–246. https://doi.org/10.1145/3394171.3413661
Niemeyer, J., Rottensteiner, F., & Soergel, U. (2014). Contextual classification of lidar data and building object detection in urban areas. ISPRS Journal of Photogrammetry and Remote Sensing, 87, 152–165. https://doi.org/10.1016/j.isprsjprs.2013.11.001
Pan, Z., Zhang, N., Gao, W., Liu, S., & Li, G. (2025). Point cloud semantic segmentation with sparse and inhomogeneous annotations. Proceedings of the AAAI Conference on Artificial Intelligence, 39(6), 6354–6362. https://doi.org/10.1609/aaai.v39i6.32680
Panday, U. S., & Gerke, M. (2012). Fitting of parametric building models to oblique aerial images. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XXXVIII-4/W19, 233–238. https://doi.org/10.5194/isprsarchives-xxxviii-4-w19-233-2011
Thomas, H., Qi, C. R., Deschaud, J.-E., Marcotegui, B., Goulette, F., & Guibas, L. J. (2019). KPConv: Flexible and deformable convolution for point clouds. https://doi.org/10.48550/ARXIV.1904.08889
Varney, N., Asari, V. K., & Graehling, Q. (2020). DALES: A large-scale aerial LiDAR data set for semantic segmentation. https://doi.org/10.48550/ARXIV.2004.11985
Vijaywargiya, J., & Ramiya, A. M. (2025). Thiruvananthapuram aerial LiDAR dataset (TALD): A benchmark for complex urban point cloud. IEEE Access, 13, 42350–42363. https://doi.org/10.1109/access.2025.3546628
Wu, X., Jiang, L., Wang, P.-S., Liu, Z., Liu, X., Qiao, Y., Ouyang, W., He, T., & Zhao, H. (2023). Point transformer V3: Simpler, faster, stronger. https://doi.org/10.48550/ARXIV.2312.10035
Xu, C., Wu, B., Wang, Z., Zhan, W., Vajda, P., Keutzer, K., & Tomizuka, M. (2020). SqueezeSegV3: Spatially-adaptive convolution for efficient point-cloud segmentation. https://doi.org/10.48550/ARXIV.2004.01803
Ye, Z., Xu, Y., Huang, R., Tong, X., Li, X., Liu, X., Luan, K., Hoegner, L., & Stilla, U. (2020). LASDU: A large-scale aerial LiDAR dataset for semantic labeling in dense urban areas. ISPRS International Journal of Geo-Information, 9(7), 450. https://doi.org/10.3390/ijgi9070450
Zachar, P., Bakuła, K., & Ostrowski, W. (2023). CENAGIS-ALS benchmark - new proposal for dense ALS benchmark based on the review of datasets and benchmarks for 3D point cloud segmentation. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-1/W3-2023, 227–234. https://doi.org/10.5194/isprs-archives-xlviii-1-w3-2023-227-2023
Zhao, H., Jiang, L., Jia, J., Torr, P., & Koltun, V. (2020). Point transformer. https://doi.org/10.48550/ARXIV.2012.09164
Zolanvari, I., Ruano, S., Rana, A., Cummins, A., Smolic, A., Da Silva, R., & Rahbar, M. (2019, September). DublinCity: Annotated LiDAR point cloud and its applications.